Obsidian Stats logoObsidian Stats

Local Dictation

by Alexander Brittain
5
4
3
2
1
Score: 35/100
New Plugin
Install Code

Description

This plugin has not been manually reviewed by Obsidian staff. Dictate notes with Whisper or Cohere Transcribe; clean up with a local Ollama model. Private, on-device speech-to-text for Obsidian.

Reviews

No reviews yet.

Stats

stars
downloads
0
forks
0
days
NaN
days
NaN
days
0
total PRs
0
open PRs
0
closed PRs
0
merged PRs
0
total issues
0
open issues
0
closed issues
0
commits

Latest Version

Invalid date

Changelog

See all version on GitHub

README file from

Github

Local Dictation

Private, on-device speech-to-text for Obsidian. Dictate notes with Whisper or Cohere Transcribe; clean up with a local Ollama model.

GitHub release GitHub stars License: MIT

✨ Features

  • Cohere Transcribe — a Hugging Face Open ASR Leaderboard-topping engine, running locally.
  • Whisper — mature offline transcription with a range of size/speed options.
  • Silero v6 VAD — enterprise-grade neural voice activity detection for real-time speech boundary detection.
  • Optional Ollama LLM cleanup — polish dictated text with a local LLM.
  • One-click model management — browse, download, and remove models from inside the plugin.
  • Hardware acceleration — Metal on macOS, CUDA on Linux/Windows (Turing-or-newer NVIDIA GPUs).
  • Private and offline — transcription stays on-device. No cloud, no telemetry, no account. Only model downloads need a network.

💻 Platform Support

Platform Hardware Acceleration
macOS Metal for Whisper (automatic via system frameworks).
Linux CUDA for Whisper and Cohere on Turing-or-newer NVIDIA GPUs. Flatpak installs need a GPU setup step.
Windows CUDA for Whisper and Cohere on Turing-or-newer NVIDIA GPUs — see Windows CUDA setup.

CPU works everywhere with no extra dependencies. CUDA acceleration requires an RTX 20-series / GTX 16-series or newer GPU with a driver compatible with CUDA 12.9; Cohere on CUDA also needs cuDNN 9 (falls back to CPU without it). Full details in Platform Runtime Dependencies.

🚀 Quick Start

Install Local Dictation from Obsidian's Community Plugins. On first run, a setup wizard walks you through downloading the speech engine and picking a transcription model — that's the easiest path.

If you'd rather do it manually, open Settings → Local Dictation: install the sidecar from the sidecar section, then click Manage models to download a model. Either way, once setup finishes you can dictate from the mic in the ribbon, or bind a hotkey to the Local Dictation: Toggle dictation command.

Where things live:

  • Sidecar binary: inside the plugin folder, under .obsidian/plugins/local-dictation/bin/.
  • Models: stored outside your vault, in your user data directory, so they aren't duplicated per-vault:
    • Windows: %LOCALAPPDATA%\obsidian-local-stt\models
    • macOS: ~/Library/Application Support/obsidian-local-stt/models
    • Linux: ~/.local/share/obsidian-local-stt/models

🔒 Privacy

Local Dictation is built to be private. Your audio and your notes never leave your machine. There is no account, no cloud service, no telemetry, and no background network traffic.

To make local transcription work, the plugin does a few things:

  • Installs a helper program. A small native "sidecar" is downloaded once from this repository's GitHub Releases and stored inside the plugin's folder. The plugin runs this helper locally to do the actual transcription.
  • Stores model files on disk. Whisper and voice-activity models are cached outside your vault so they aren't duplicated per-vault. You can browse and remove them from the plugin's model manager.
  • Uses the network only for downloads. The sidecar archive and model files are fetched from their official sources on demand. Nothing else is sent anywhere.

🤝 Contributing

See CONTRIBUTING.md for prerequisites, setup, scripts, branching conventions, PR workflow, and architecture overview.

📄 License

MIT. See LICENSE.

Sponsors